Using an AGROVOC-based ontology for the
description of learning resources on organic agriculture
Salvador Sánchez-Alonso and Miguel-Angel Sicilia
Information Engineering Research Unit
Computer Science Dept., University of Alcalá
Ctra. Barcelona km. 33.6 – 28871 Alcalá de Henares (Madrid), Spain
{salvador.sanchez, msicilia}@uah.es
Abstract. Education is a critical requirement for the development of sustainable
agriculture. Learning resources available through the Web can be described
with metadata for enhanced availability. The provision of semantic metadata
describing these resources further facilitates retrieval and selection of resources
based on richer annotations that use ontologies. This paper sketches the
potential use of ontologies related to organic agriculture for the description of
learning resources.
Keywords: Thesauri, ontology, AGROVOC, organic agriculture, learning
objects.
1
Introduction
Organic agriculture is a form of agriculture whose main objective is obtaining food
efficiently while respecting the environment and preserving Earth’s natural fertility,
which is attained through the optimization of the resources available as well as
avoiding synthetic pesticides and fertilizers. Although organic farming is nowadays
widespread in most developed countries, the promotion of the ecological practices in
agriculture requires much effort in terms of education. Different institutions and
organizations provide educational resources on the topic, some of them openly
available through the Web. However, locating those resources with conventional
search engines is complicated, mainly due to noise in the results of common input
terms. Learning object repositories provide an alternative –which can be seen as an
extension–, which enables more relevant results at the cost of developing (and packing
together with the learning objects) a few metadata records. The IEEE LOM standard1
can be used to provide metadata to learning resources thus facilitating their retrieval.
There exists, however, an additional extension that would provide richer means of
browsing, navigating and searching for educational resources: the use of formal
annotation based on (formal) ontologies. This approach is specially suited to the field
of agriculture, since the large and mature thesaurus AGROVOC2, a vocabulary
1
2
http://ltsc.ieee.org/wg12/
http://www.fao.org/agrovoc/
covering the terminology of subject fields in agriculture, forestry, fisheries, food and
related domains, is widely used in practice, which represents a degree of consensus
regarding terminology.
Moving to a Semantic Web practice requires essentially three elements: (i) the
elaboration of a formal ontology from the thesaurus, (ii) the provision of specialized
semantic search software and (iii) the change in current indexing practices, from the
traditional use of thesauri to the new semantic annotation. This paper focuses on
element (i), providing a tentative mapping technique for the specifics of AGROVOC.
The technique is targeted to the annotation of learning objects, reusing the effort
carried out in ontologies for learning resources (Sicilia et al., 2004).
AGROVOC is a structured, controlled vocabulary used for indexing and retrieving
data in agricultural information systems. It consists of organized terms (in different
languages) covering not only the terminology of agriculture, but also terms in forestry,
fisheries, food and other related domains. These terms are used to unambiguously
identify resources. Indeed, the knowledge contained in the vocabulary allows
standardizing indexing processes, making searching simpler and more efficient.
As in other thesauri, terms are related in AGROVOC, but the kind of relationships
supported is very limited:
- Broader term (BT) relationships link a general term to other(s) more specific.
Thus, the concept Soil is BT related to more general terms such as Land
cover.
- Narrower term (NT) relationships represent the opposite of BT. The concept
Soil is NT related to the three more specific concepts Top soil,
Rhizosphere and Subsoil.
- Related term (RT) relationships link any two concepts holding a non
hierarchical relationship. The term Fish, for instance, is RT related to terms
as varied as Foods, Perishable products, Seafoods, Fresh
products or Postmortem changes.
Other less relevant relationships in AGROVOC are: Is Referenced in Scope Note
(SNX), Scope Note Reference (SNR), See (SEE), Seen for (SF), Use (USE) and Used
for (UF).
AGROVOC was developed by FAO (Food and Agriculture Organization of the
United Nations) and the Commission of the European Communities in the early
1980s, being updated since then on a regular basis and extensively used for indexing
and retrieving data in agricultural information systems. Understanding that similar
efforts such as the European GEMET exist (de Lavieter, 1995), the number of terms
included (up to 40.000 terms per language), the continuous support and funding from
FAO, as well as its generally acceptance, number of active partners and widespread
use, make of AGROVOC an outstanding resource in the field of vocabularies and
certainly a point of reference for the subject fields covered.
The maturity of AGROVOC makes it a good candidate to become a point of
departure for an effort of formalization. In fact, this has already been approached; see
for example (Soergel et al., 2004). However, this paper focuses on semantic
annotation for learning resources, which would be a particular application of the
development of an AGROVOC-based ontology.
The rest of this paper is structured as follows. Section 2 reviews previous work in
the field, before section 3 exemplifies the extension of AGROVOC to formal ontology
focusing on a particular aspect of organic agriculture, fertilization, which is covered
by the thesaurus. Then, section 4 sketches some guidelines on how annotation of
learning resources on organic farming can be annotating by making use of the
ontology. Finally, conclusions and outlook are provided in section 5.
2
Timeline for the creation of an AGROVOC ontology
From 2003, several efforts have attempted to convert the AGROVOC thesaurus
into an ontology. And even though the full conversion of AGROVOC to a formal
ontology is currently ongoing work, probably due to the massive effort necessary to
properly address this issue, it is fair to recognize that some interesting work has been
done during the years.
In 2003, the thesaurus was converted into an RDFS file. Later on, several efforts
coordinated to redesign the traditional thesaurus into an ontology-based relational
database available to download. Subsequent efforts used AGROVOC in the design of
ontologies specific to given domains, such as food safety (Lauser, 2001), or
bibliographic information on agriculture (Wildemann, Salokhe and Keizer, 2004),
among others. In a more general domain, Kashyap (1999) proposed an approach for
designing an ontology for information retrieval based on databases’ schemas and a
collection of queries that are of interest to the users. However, advanced semantic
inference is not considered an issue and thus not dealt with in depth, as the main goal
is to reduce the involvement of domain experts. Closer to AGROVOC, an interesting
effort by Gangemi et al. (2002) aims at building a fishery ontology by reengineering
and integrating the fisheries terminologies from the several systems, one of which is
AGROVOC. On the other hand, Fisseha, Liang, and Keizer (2003) sketched some
ideas to transform AGROVOC to an ontology. This was an important move towards a
proper ontology of AGROVOC, but it does not address important issues such as
providing “intelligent behind-the-scenes support for query expansion” (in their own
words), or creating the complete inventory of domain-relevant entity types and
relationship types, to name a few.
In spite of the several ongoing works on developing agriculture ontologies, there
are no reports on the development of specific schemas for the annotation of learning
resources for agriculture education
3
Converting AGROVOC to ontological languages: the case of
fertilization
At present, AGROVOC contains close to 30,000 descriptors and more than 10,000
non-descriptors called synonyms (terms which help the user to find a specific
descriptor). Given the size and scope of AGROVOC, any ontology created from it
will likely include thousands of concepts and relationships, and consequently will not
be feasible to manage. Here we will focus on the concrete case of fertilization in
organic agriculture as a simple application of a part of the thesaurus.
Fertilization, understood as “any aspect of the use of fertilizers to improve crop
growth and soil fertility”, is a central concept in organic agriculture. It is a fact that no
farming exploitation can be considered –and consequently certified as– organic,
unless a strict use of fertilizers and fertilizing techniques, specific to organic
agriculture, is followed. In Europe, for instance, regulation on organic certification
enforces fertilizers and soil conditioners to be composed only of substances listed,
which particularly involves varied cultivation practices, and the rigorous limitation on
the use of non-synthetic fertilizers. In AGROVOC, the term Fertilization3 is
considered a physiological concept, so the AGROVOC term defining the application
of fertilizers to soil or plants that should be used is, instead, Fertilizer
application. This latter term is related to (RT) Fertilizers, a concept which
includes any natural or synthetic compound spread on the plant or worked into soil to
increase the plant’s natural capacity to grow. A particular kind of fertilizers is
Biofertilizers, a NT related concept describing “any naturally occurring
organic substances applied to soil for the purpose of maintaining or improving
fertility”. As Biofertilizers are exclusively composed of natural substances
such as animal manures, composts, nitrogen fixing bacteria and mycorrhizae, it is
suitable (from the point of view of current regulations) for organic farming practices.
On the contrary, other types of NT related Fertilizers such as Nitrogen
fertilizers and Phosphate fertilizers —and their derivatives
Superphosphate and Rock phosphate— are generally prohibited for organic
farming. This is because they have been named as factors in the over enrichment of
surface waters (excessive nutrients in ponds and lakes cause over growth of algae) and
phosphate accumulation in subterranean waters (extremely soluble nitrogen in its
nitrate form not sufficiently absorbed by plants can leach into groundwater). Others
types of Fertilizers (i.e. NT related to this term) are Biofertilizers, Calcium
fertilizers, Inorganic fertilizers, Liquid gas fertilizers or
Organic fertilizers, among others (see Figure 1). Figure 1 shows a fragment
of AGROVOC with NT/BT relations interpreted as class/subclass.
3
To enhance readability, AGROVOC terms are in courier font.
Figure 1. A fragment of the terms in AGROVOC related to fertilization.
An interesting term in AGROVOC is Fertilizer combinations. It
represents any fertilizer mixture with agents such as herbicides, plant growth
substances or pesticides. This is obviously not suitable for organic agriculture
according to the definition provided at the beginning of this article. At the moment,
this is one of the many terms related to Fertilizers (i.e. Fertilizer
combinations is RT related to Fertilizers). Another interesting concept to
note is that of Soil pollution, a general concept RT related to Soil
degradation which is not certainly applicable to given types of Fertilizers
such as Biological fertilizers, a powerful reason to better profile the
relationship holding between Fertilizers and Soil Pollution.
Current relationships in AGROVOC suggest the following:
- All kinds of Fertilizers can be involved in Soil degradation,
a concept RT related to Soil Pollution which is in turn related to
Fertilizers.
- Any instance of Fertilizers can be part of Fertilizers
combinations, which is probably false for some kinds of
Biological fertilizers.
Figure 1 shows the abovementioned concepts and the relations holding between
them as a previous step for its formulation and clarification in ontological terms.
In AGROVOC, four relationships showing the hierarchical links between terms exist:
BT (broader term), NT (narrower term), RT (related term) and UF (non-descriptor).
These relationships are specific to thesauri and, as such, can not be considered a
limitation or disadvantage. However, an ontology takes this conceptual framework one
step further by structuring the terms more formally, and by providing richer
relationships between concepts than what is currently provided in thesauri. An
AGROVOC ontology should ideally clarify the hierarchies defined to better prepare
the knowledge base for inferences. Thus, a number of basic actions should be taken
such as:
- Creating two new subconcepts of Fertilizer applications namely
General
fertilizer
applications
and
Specific
-
-
-
fertilizer
applications.
General
fertilizer
applications would give support to fertilizer techniques suitable for all
kinds of farming practices (organic and non organic). These techniques will
likely include the utilization of permitted Fertilizers and will explicitly avoid
non permitted compounds. On the other hand, Specific fertilizer
applications, would provide support to techniques suitable for specific
forms of agriculture. This can be subsequently divided into Organic
agriculture
fertilizers
applications (for organic
agriculture techniques), extensive agriculture techniques, etc.
A subclass of Fertilizers, should be created to give support to the
concept of Organic aggriculture-suitable fertilizer. Let us name this term
Organic agriculture-suitable fertilizers.
Remove the relationship between Fertilizer application and
Fertilizers, as in its current for it provides room for errors in selecting
the appropriate resource if the selection is based on the ontology knowledge.
A good solution (but more complex than current situation) is to link
Organic
agriculture
fertilizer
applications to
Organic agriculture-suitable fertilizers through an
ontology property named makesUseOf.
Link Soil pollution only to those fertilizers which might be cause of
pollution; let us name them Pollutant-prone fertilizers. In this
manner, the rest of the different types of Fertilizers will no longer be
able to be associated to Soil degradation through Soil
pollution as this relationship will not hold to all the Fertilizers but
only to those pollution-prone.
Figure 2 summarizes all the above discussion.
Figure 2. Fragment of an AGROVOC-based ontology for fertilization.
4
Case study: annotating resources on fertilization
Learning resources on organic agriculture may be targeted to different kinds of
learners. Here we will provide two cases that illustrate such diversity of needs. The
first will be a scientific paper while the second is a resource oriented to dissemination
of technical practices.
Annotating technical resources
In general, any digital resource can be explicitly declared as an instance of a
LearningObject class – or any of its subclasses. Then, there is a family of
properties (and subproperties) starting from a generic about that connects them to
any other concept. The about predicates play the same role of a “keywords” marker,
but it can be specialized. One important educational issue is the production of organic
agriculture fertilizers. However, the notion of what an OA fertilizer is can be defined
in several ways. A possible definition can be provided by the following two SWRL
rules:
NaturalOrganicFertilizer(?x) → OrganicAgricultureFertilizer(?x)
agv_Fertilizer(?x)
∧
OrganicAgricultureStandard(?y)
OrganicAgricultureFertilizer(?x)
defines(?y,
?x)
∧
→
That is, natural organic fertilizers are suitable to OA but also any fertilizer defined as
such by standards or recommendations.
Composting is one of the most widespread of these production techniques. The
“Compost happens!” tutorial (CHT4) is an online resource with a common sequential
structure.
This
resource
can
be
declared
as
an
instance
TutorialLearningObject(cht).
The tutorial can be annotated as
describes(cht,
composting)
where
FertilizerProductionTechnique(composting).
A user searching for techniques for producing organic agriculture fertilizers
(anything about OAFertilizerProductionTechnique) will match the
resource since produces(composting, compost) and compost will be
classified as OA, for example, because it is defined by the BSI_PAS_100 standard.
Annotating scientific literature
Palomäki et al (2002) reported a typical comparative study of organic versus
traditional techniques, in this case for the particular kind of Elsanta strawberries. This
kind of resources considered as LO are considered ScientificReportLO, and
they can be annotated with respect to the concrete research methods employed.
A user might first search for reports regarding StrawberryPlantType that
report on Experiments (specified in the researchMethodUsed property). The
report mentioned will be of the kind ComparativeExperiment. A more detailed
specification may specify as measuredVariable fruitProductivity. This
kind of measures can be used for a systematization of research evidence, but it also
can be retrieved as supplementary material for expositive or tutorial learning objects,
which are related
5
Conclusions and outlook
An example on how an ontology fragment derived from AGROVOC can be used to
annotate resources has been provided. Further work will deal with advancing the
conversion of AGROVOC to formal ontology, and with the provision of tools and
improved techniques for semantic annotation specific to learning resources.
4
http://www.compostinfo.com/main/intro.htm
Acknowledgements
This work has been supported by project LUISA (Learning Content Management
System Using Innovative Semantic Web Services Architecture), code
FP6−2004−IST−4 027149.
References
Council Regulation (EEC) No 2092/91 of 24 June 1991 on organic production of agricultural
products and indications referring thereto on agricultural products and foodstuffs.
Fisseha, F., Liang, A. and Keizer, J. (2003). Reengineering AGROVOC to Ontologies
Step towards better semantic structure. In Proceedings of the NKOS Workshop, Rice
University, Houston, Texas, USA.
Gangemi, A., Fisseha, F., Pettman, I. Jand Keizer, J. (2002). Building an Integrated Formal
Ontology for Semantic Interoperability in the Fishery Domain. In Proceedings of the First
International Semantic Web Conference, ISWC 2002.
Kashyap V. (1999) Design and creation of ontologies for environmental information retrieval.
In Proceedings of the 12th Workshop on Knowledge Acquisition, Modeling and
Management (KAW'99), Banff, Canada.
Lauser, B. (2001). From thesauri to Ontologies: A short case study in the food safety area in
how ontologies are more powerful than thesauri - From thesauri to RDFS to OWL. In
proceedings of the ECDL 2001 OAI Workshop.
de Lavieter, L. (Ed.) (1995). Multilingual Environmental Thesaurus. Part 1, English; Part 2,
Français; Part 3, Deutsch; Part 4, Nederlands; Part 5, Italiano; Part 6, Norsk; Part 7, Dansk;
Part 8, Español. NBOI, Nederlandse Bureau voor Onderzoek Informatie / EEA-TF European Environment Agency - Task Force, Amsterdam, November 1995, pp. (English) vi
+ A-78; B-112; C-56; D-199, total 445.
Palomäki, V., Mansikka-aho, A.-.M. and Etelämäki, M. 2002. ORGANIC FERTILIZATION
AND CULTIVATION TECHNIQUE OF STRAWBERRY GROWN IN GREENHOUSE.
Acta Hort. (ISHS) 567:597-599 http://www.actahort.org/books/567/567_129.htm
Sicilia, M.A., García, E., Sánchez-Alonso, S. and Rodríguez, E. 2004. Describing learning
object types in ontological structures: towards specialized pedagogical selection. In
Proceedings of ED-MEDIA 2004 - World conference on educational multimedia,
hypermedia and telecommunications, pp. 2093-2097.
Soergel, D., Lauser, B., Liang, A., Fisseha, F., Keizer, J. and Katz, S. (2004). Reengineering
thesauri for new applications: the AGROVOC example. Journal of Digital Information,
4(4).
Wildemann, T., Salokhe, G. and Keizer, J. (2004). Applying new trends to the management of
bibliographic information on agriculture. Zeitschrift für Agrarinformatik, 12(1), 9-13.